410 likes | 561 Views
The Emergence of Networking Abstractions and Techniques in TinyOS. Title. The Emergence. of Networking Abstractions and Techniques. in TinyOS. Tranditional OS. Multi thread Blocking I/O Model Kernel / User area No power limitaion. Sensor Network OS. Power limitation Small memory
E N D
The Emergence of Networking Abstractions and Techniquesin TinyOS
Title The Emergence of Networking Abstractions and Techniques in TinyOS
Tranditional OS • Multi thread • Blocking I/O Model • Kernel / User area • No power limitaion
Sensor Network OS • Power limitation • Small memory • Restricted H/W • Wireless Commucation
TinyOS • Origin from Smart Dust project of US. Berkely • Embedded • Componente-based • Completely non-blocking
TinyOS: Design • Current and likely future design • Diverse implementation • Address the specific and unusual challenges
TinyOS: H/W • rene mote • Released in 2000 • 4MHz, 8kB flash, 0.5kB RAM, 10kBaud
TinyOS: H/W • mica • Released in 2002 • 4MHz, 128kB flash, 4kB RAM, 40kBaud
TinyOS: H/W • mica2 • Released in 2003 • 7MHz, 128kB flash, 4kB RAM, 40kBaud
TinyOS: H/W • iMote • Released in 2003 • 12MHz, 512kB flash, 64kB RAM, 460kBaud
TinyOS: Application • Habitat Monitoring • Great Duck Island • Vineyard in British Columbia • Keep energy consumption low
TinyOS: Application • Shooter Localization • Origin of bullet in an urban setting • A high sample rate and fine-grain time sync
TinyOS: Application • Pursuer-Evader • Mote localization and advanced route
Network in TinyOS • Low bandwidth • Consumes majority of the active power • Issues • Discovery • Routing • Reliability • Congestion control
Single hop: Active Messages • rene Gray Components abstract hardware
Single hop: Active Messages • mica Gray Components abstract hardware
Single hop: Active Messages • mica2 Gray Components abstract hardware
Single hop: Active Messages • mica S-MAC Gray Components abstract hardware
Multi hop • Tree-based collection • Route or aggregate data to endpoint • Intra-networking routing • Data is transferred between in-network endpoints • And disseminate to entire region
Tree-Based Routing • Based on • A parent node ID • Hop-count or depth from the tree root • A node routes a packet by transmitting it with the parent • Until the packet reaches the root of the tree • AMROUTE, BLess, Surge, mh6, MultiHopRouter
Intra-network Routing • DSDV, AODV • Unicast routing to specific endpoints • GPSR • Node’s geographical locaion • Intra-network routing is uncommon in TinyOS • Except Pursuer-Evader
Broadcast and Epidemic Protocols • Broadcast • For reconfigure filter setting or radio transmit power • For distribute new version of TinyOS programs • Simple flooding protocol • Common, easily implemented • Epidemic algorithm • Epidemic only transmits when needed • Local suppression mechanisms can reduce redundant transmission: saving energy
Network Service: Power Management • TinyOS manages power management • Each service can be stopped • HPLPowerManagment Component put the processor into the lowest-power mode • TinyOS timer service can function in the extremely-low-power, power-save mode • Cross-layer control at a very low-level • HPLPowerManagement can switch the processor into various lower-power modes.
Time Synchronization • For combine a set of coincident readings from different location, need to establish the temporal consistency of data • TDMA-style media access protocols for slot coordination • Power-aware approaches to communication scheduling • Low-level time sync were unsuccessful • Get and set system time and time stamp message
General Abstraction • The AM abstraction • Stable since the earliest TinyOS • Tree-based routing(the Send and Intercept interfaces) • Route(Berkeley), HSN, AODV, DSDV() use this interface
Specialized Abstraction • Power management and Time synchronization • General abstractions of these services are very hard to get right • Requirements of applications vary dramitically • Accurate to within a few milliseconds with a small set of other nodes • Globally synchronized clock that is much less accruate
In-Flux Abstractions • Commonly found but changing between applications and H/W version • Epidemic propagation • Radio MAC • Channel activity, the use of control packet per data packet, backoff, link estimation
Absent Abstractions • Expected to find in TinyOS but that were absent in the code base • Distributed cluster formation • Incoming(receive) queues
Communication Scheduling and Snooping • Communication scheduling • To disabling the radio expect druing pre-arranged time • Snooping • Receiving packets that might not even be destined for a node • To acquire network neighborhood
Cross-Layer Control • Routing stacks share network neighborhood information between link state(low layer) and network layer(higher layer) • Avoiding duplicating data • Conserving RAM • Provision by the network stack of low-level information to a higher level • Avoid unnecessary communication
Static Resource Allocation • Allow buffer for the network, sensor, UART and the other OS services at complie-time • RAM is valuable in TinyOS
EmNets vs. the Internet • Networking Abstractions in sensor network differ from tranditional Internet abstractions • Resource constraints • Very different set of goal and principles • Sensor networks are • Homogeneous system • Application-specific and collaborative perpose • Every node is both a sensor and a router
Conclusion • Classify the most prominent abstractions in TinyOS • General, Special, In-flux, Absent • Find Several techqiunces • Closs-layer control • Static resource allocation • WSNs is drivend by • Power management • Limited resource • Real-time conistraints